In the past decade, processor speed has become significantly faster than memory speed. Small, fast cache memories are designed to overcome this discrepancy, but they are only effective when programs exhibit data locality. In this article, we present compiler optimizations to improve data locality based on a simple yet accurate cost model. The model computes both temporal and spatial reuse of cache lines to find desirable loop organizations. The cost model drives the application of compound transformations consisting of loop permutation, loop fusion, loop distribution, and loop reversal. We demonstrate that these program transformations are useful for optimizing many programs. To validate our optimization strategy, we implemented our algorithms and ran experiments on a large collection of scientific programs and kernels. Experiments illustrate that for kernels our model and algorithm can select and achieve the best loop structure for a nest. For over 30 complete applications, we executed the original and transformed versions and simulated cache hit rates. We collected statistics about the inherent characteristics of these programs and our ability to improve their data locality. To our knowledge, these studies are the first of such breadth and depth. We found performance improvements were difficult to achieve because benchmark programs typically have high hit rates even for small data caches; however, our optimizations significantly improved several programs.

No abstract

Compiler transformations can significantly improve data locality for many scientific programs. In this paper, we show iterative solvers for partial differential equations (PDEs) in three dimensions require new compiler optimizations not needed for 2D codes, since reuse along the third dimension cannot fit in cachefor larger problem sizes. Tiling is a program transformation compilers can apply to capture this reuse, but successful application of tiling requires selection of non-conflicting tiles and/or padding array dimensions to eliminate conflicts. We present new algorithms and cost models for selecting tiling shapes and array pads. We explain why tiling is rarely needed for 2D PDE solvers, but can be helpful for 3D stencil codes. Experimental results show tiling 3D codes can reduce miss rates and achieve performance improvements of 17-121% for key scientific kernels, including a 27% average improvement for the key computational loop nest in the SPEC/NAS benchmark MGRID.

Abstract. This paper presents an unbalanced tree search (UTS) benchmark designed to evaluate the performance and ease of programming for parallel applications requiring dynamic load balancing. We describe algorithms for building a variety of unbalanced search trees to simulate different forms of load imbalance. We created versions of UTS in two parallel languages, OpenMP and Unified Parallel C (UPC), using work stealing as the mechanism for reducing load imbalance. We benchmarked the performance of UTS on various parallel architectures, including sharedmemory systems and PC clusters. We found it simple to implement UTS in both UPC and OpenMP, due to UPC's shared-memory abstractions. Results show that both UPC and OpenMP can support efficient dynamic load balancing on shared-memory architectures. However, UPC cannot alleviate the underlying communication costs of distributed-memory systems. Since dynamic load balancing requires intensive communication, performance portability remains difficult for applications such as UTS and performance degrades on PC clusters. By varying key work stealing parameters, we expose important tradeoffs between the granularity of load balance, the degree of parallelism, and communication costs.

As the speed of mass spectrometers, sophistication of sample fractionation, and complexity of experimental designs increase, the volume of tandem mass spectra requiring reliable automated analysis continues to grow. Software tools that quickly, effectively, and robustly determine the peptide associated with each spectrum with high confidence are sorely needed. Currently available tools that postprocess the output of sequence-database search engines use three techniques to distinguish the correct peptide identifications from the incorrect: statistical significance re-estimation, supervised machine learning scoring and prediction, and combining or merging of search engine results. We present a unifying framework that encompasses each of these techniques in a single model-free machinelearning framework that can be trained in an unsupervised manner. The predictor is trained on the fly for each new set of search results without user intervention, making it robust for different instruments, search engines, and search engine parameters. We demonstrate the performance of the technique using mixtures of known proteins and by using shuffled databases to estimate false discovery rates, from data acquired on three different instruments with two different ionization technologies. We show that this approach outperforms machine-learning techniques applied to a single search engine's output, and demonstrate that combining search engine results provides additional benefit. We show that the performance of the commercial Mascot tool can be bested by the machine-learning combination of two open-source tools X!Tandem and OMSSA, but that the use of all three search engines boosts performance further still. The Peptide identification Arbiter by Machine Learning (PepArML) unsupervised, model-free, combining framework can be easily extended to support an arbitrary number of additional searches, search engines, or specialized peptide-spectrum match metrics for each spectrum data set. PepArML is open-source and is available from http://peparml.sourceforge.net.